1: 

2: 

3: \documentclass[12pt,preprint]{aastex}

4: 

5: \def\beq{\begin{eqnarray}}

6: \def\eeq{\end{eqnarray}}

7: 

8: \begin{document}

9: 

10: \title{CONSTRAINTS ON THE MASS ACCRETION RATE OF NEUTRINO-COOLED DISKS IN GAMMA-RAY BURSTS}

11: 

12: 

13: \author{Tong Liu, Wei-Min Gu, Li Xue, Shan-Shan Weng, and Ju-Fu Lu*}

14: 

15: \affil{Department of Physics

16: and Institute of Theoretical Physics and Astrophysics, \\

17: Xiamen University, Xiamen, Fujian 361005, China}

18: 

19: \email{*lujf@xmu.edu.cn}

20: 

21: \begin{abstract}

22:  We present a unified description of all the three known classes of optically

23: thick accretion disks around black holes, namely Shakura-Sunyaev

24: disks, slim disks, and neutrino-dominated accretion flows (NDAFs).

25: It is found that NDAFs have both a maximal and a minimal possible

26: mass accretion rate at their each radius. This may be suggestive of

27: an interpretation for the origin of X-ray flares observed in

28: gamma-ray bursts.

29: \end{abstract}

30: 

31: \keywords{accretion, accretion disks - black hole physics - gamma

32: rays: bursts - neutrinos}

33: 

34: \section{Introduction}

35: 

36: Recently, Gu \& Lu (2007) addressed a theoretical problem regarding

37: the slim accretion disk model in the fundamental sense. In this

38: model (e.g., Abramowicz et al. 1988; Kato et al. 1998, p.242), in

39: dealing with the vertical hydrostatic equilibrium of the disk, the

40: well-known pseudo-Newtonian potential of Paczy\'nski \& Wiita

41: (1980), \beq\ \psi (r,z) = - \frac {G M}{\sqrt{r^2+z^2}-r_{\rm g}},\

42: \eeq was approximated in the form suggested by H\={o}shi (1977),

43: i.e.,\beq\ \psi (r,z) = \psi (r,0) + \Omega_{\rm K} ^2 z^2/2,\ \eeq

44: where $r$ and $z$ are cylindrical coordinates, $M$ is the mass of

45: the central black hole, $r_{\rm g} \equiv 2 G M / c^{2} $ is the

46: Schwarzschild radius, and $\Omega_{\rm K} = (GM /r)^{1/2} /(r -

47: r_{\rm g})$ is the Keplerian angular velocity. As shown by Gu \& Lu

48: (2007), equation (2) is valid only for geometrically thin disks such

49: as Shakura-Sunyaev disks (SSDs, Shakura \& Sunyaev 1973), i.e., with

50: $H \ll r$, where $H$ is the half-thickness of the disk; and this

51: equation is invalid for slim disks that may have $H \lesssim r$ or

52: $H \sim r$, because in this case it would greatly magnify the

53: gravitational force in the vertical direction of the disk, $\partial

54: \psi /

55: \partial z$. When the explicit form of Paczy\'nski-Wiita potential,

56: equation (1), is used to calculate the vertical gravitational force,

57: the relationship that is obtained with equation (2) and is

58: applicable only for geometrically thin disks, i.e., $c_{\rm s} /

59: \Omega_{\rm K} H = {\rm constant}$, where $c_{\rm s}$ is the sound

60: speed, does not hold for slim disks. Accordingly and more seriously,

61: it is found that slim disks cannot exist at large radii of black

62: hole accretion flows with large accretion rates, and only the inner

63: regions of these flows can possibly take the form of slim disks

64: provided accretion rates are effectively reduced by outflows from

65: the outer regions.

66: 

67: Slim disk are optically very thick in the vertical direction, such

68: that photons are sufficiently trapped within the disk and advected

69: along with the disk matter into the black hole. In this paper, we

70: extend the work of Gu \& Lu (2007) into another class of accretion

71: disks, namely neutrino-cooled accretion disks (or neutrino-dominated

72: accretion flows, NDAFs) around stellar-mass black holes. These disks

73: are known to be plausible candidates for the central engines of

74: gamma-ray bursts (Popham et al. 1999; Narayan et al. 2001; Kohri \&

75: Mineshige 2002; Di Matteo et al. 2002; Kohri et al. 2005; Lee et al.

76: 2005; Gu et al. 2006; Chen \& Beloborodov 2007; Liu et al. 2007;

77: Kawanaka \& Mineshige 2007; Janiuk et al. 2007). They can be

78: regarded as an even more extreme case of optically thick disks,

79: since their density and temperature are so high ($\rho \sim

80: 10^{10}$g cm$^{-3}$, $T \sim 10^{10} {\rm K}$) that photons are

81: completely trapped and only energetic neutrinos are emitted away.

82: 

83: 

84: \section{Equations}

85: 

86: 

87: NDAFs may have $H$ comparable to $r$ (e.g., Popham et al. 1999; Chen

88: \& Beloborodov 2007; Janiuk et al. 2007), so their hydrodynamics and

89: thermodynamics are expected to be similar to those of slim disks. We

90: write the continuity, radial momentum, angular momentum, and energy

91: equations of NDAFs in the formalism of Gu \& Lu (2007) for slim

92: disks, where equation (1) was used to integrate the vertical

93: hydrostatic equilibrium equation and the simple relation $c_{\rm s}

94: / \Omega_{\rm K} H = {\rm constant}$ was abandoned: \beq {\dot M} =

95: -2 \pi r \Sigma v_{r} = {\rm constant} \ , \eeq \beq v_{r}\frac{{\rm

96: d} v_{r}}{{\rm d}{\rm ln} r} + c_{\rm s}^2 \frac{{\rm d}{\rm ln}

97: \Pi} {{\rm d}{\rm ln} r} - \Omega ^ {2} r^{2} = - \frac{\Omega_{\rm

98: K}^2 (r-r_{\rm g})^2}{\Sigma} \int_{-H}^{H}

99: \frac{\rho}{[\sqrt{1+(z/r)^2}-r_{\rm g}/r]^2 \sqrt{1+(z/r)^2}} {\rm

100: d} z \ , \eeq \beq {\dot M} (\Omega r^2-j_0) = 2 \pi \alpha r^2 \Pi

101: \ , \eeq \beq Q_{\rm vis} = Q_{\rm adv} + Q_{\nu} \ . \eeq  In these

102: equations, $\dot M$ is the mass accretion rate, $v_{r}$ is the

103: radial velocity, $\Omega$ is the angular velocity, $\Sigma =

104: \int_{-H}^{H} {\rho} {\rm d} z$ is the surface density, $\Pi =

105: \int_{-H}^{H} {p}  {\rm d} z$ is the vertically integrated pressure,

106: $\rho$ is the mass density, $p$ is the pressure, the sound speed is

107: defined as $c_{\rm s} = (\Pi/\Sigma)^{1 / 2}$, $j_0$ is an

108: integration constant representing the specific angular momentum

109: accreted by the black hole, and $\alpha$ is the Shakura-Sunyaev

110: viscosity parameter. The viscous heating rate is $Q_{\rm vis} =

111: {\dot M} \Omega ^2 f g / 2 \pi$, where $f = 1 - j/\Omega_{\rm K}

112: r^2$, $j = j_0/\omega$, $\omega = \Omega/\Omega_{\rm K}$ is assumed

113: to be a constant that is smaller than 1 (sub-Keplerian rotation),

114: and $g = - {\rm d ln} \Omega_{\rm K}/{\rm d ln} r$; the advective

115: cooling rate is $Q_{\rm adv} = \xi {\dot M} {c_{\rm s}}^2  / 2\pi

116: r^2$, with $\xi$ being a dimensionless quantity of the order of

117: unity (Kato et al. 1998, p.272). All these definitions and relations

118: are similar to those for slim disks. What is new in the case of

119: NDAFs, however, is that in the energy equation (6) there is a

120: cooling term due to neutrino radiation, $Q_{\nu}$, instead of the

121: cooling term due to photon radiation, $Q_{\rm rad}$, because of the

122: photon trapping. The neutrino cooling is expressed by a bridging

123: formula that is valid in both the neutrino optically thin and thick

124: regimes:\beq\ Q_{\nu}=\sum_{i} \frac{(7/8) {\sigma} T^4 }

125: {(3/4)[\tau_{{\nu}_i}/2+1/ \sqrt{3}+1/(3 \tau_{a,{\nu}_i})]}  \eeq

126: (Korhi et al. 2005), where $T$ is the temperature, $\tau_{{\nu}_i}$

127: is the total optical depth for neutrinos, $\tau_{a,{\nu}_i}$ is the

128: absorption optical depth for neutrinos, and the subscript $i$ runs

129: over the three species of neutrinos $\nu_{\rm e}$ , $\nu_\mu$ , and

130: $\nu_\tau$ (see, e.g., Kohri et al. [2005] and Liu et al. [2007] for

131: detailed analyses and calculations of these optical depths).

132: Accordingly, the equation of state is also different from that for

133: slim disks, it is written as \beq p = p_{\rm gas} + p_{\rm rad} +

134: p_{\rm e} + p_{\nu} ,\eeq where $p_{\rm gas}$ , $p_{\rm rad}$ ,

135: $p_{\rm e}$ , and $p_\nu$ are the gas pressure from nucleons,

136: radiation pressure of photons, degeneracy pressure of electrons, and

137: radiation pressure of neutrinos, respectively. Of these four

138: pressure components, there are only the first two for slim disks,

139: while the last two are newly appeared for NDAFs. Detailed

140: expressions for these four components are also given in, e.g., Kohri

141: et al. (2005) and Liu et al. (2007). An additional note about the

142: energy equation (6) is that, for simplicity, we ignore another

143: cooling term due to photodisintegration of $\alpha$-particles and

144: other heavier nuclei; in other words, we assume that in NDAFs all

145: heavy nuclei are already disintegrated into nucleons.

146: 

147: 

148: \section{Unified Description of SSDs, Slim Disks, and NDAFs}

149: 

150: 

151: With the procedure similar to that of Gu \& Lu (2007), thermal

152: equilibrium solutions at a certain radius can be obtained from

153: equations (3) - (6) and (8), with given constant parameters $M$,

154: $\dot M$, $\alpha$, and $j$. In our calculations we take $M=3

155: M_\odot$, $\alpha=0.1$, and $j=1.83 c r_{\rm g}$.

156: 

157: 

158:   As our main result, Figure 1 shows thermal equilibria of NDAFs

159: at each radius $r$ with corresponding accretion rates $\dot M$. The

160: left and right vertical axes are for $\dot M$ in units of the

161: Eddington accretion rate ${\dot M}_{\rm Edd} = 64 \pi GM / c

162: \kappa_{\rm es}$, where $\kappa_{\rm es} = 0.34 {\rm cm}^2 {\rm

163: g}^{-1}$ is the electron scattering opacity, and of $M_\odot$

164: s$^{-1}$ , respectively. To have a complete picture of all the known

165: classes of optically thick accretion disks around black holes, the

166: results of Gu \& Lu (2007) for SSDs and slim disks are also included

167: in the figure. It is seen that the $\dot M$-$r$ plane is divided

168: into four regions by five lines $a$, $b$, $c$, $d$, and $e$. As

169: described in Gu \& Lu (2007), the region below line $a$ is for

170: stable, photon radiation-cooled and gas pressure-supported SSDs; the

171: region between lines $a$ and $b$ is for unstable, photon radiation

172: pressure-supported but not yet advective cooling-dominated SSDs; the

173: region between lines $b$ and $c$ is for stable, advective

174: cooling-dominated and photon radiation pressure-supported slim

175: disks; and line $c$ represents a maximal possible accretion rate for

176: each radius, above which no thermal equilibrium solutions exist

177: because the viscous heating $Q_{\rm vis}$ is always larger than the

178: total cooling $(Q_{\rm adv} + Q_{\rm rad})$. What is new in the

179: figure is lines $d$ and $e$. The 'no solution' region above line $c$

180: extends upward till line $d$, above which the cooling due to

181: neutrino radiation, $Q_\nu$, becomes important or even dominant,

182: such that thermal equilibrium solutions can be established again,

183: and these are exactly NDAF solutions. Thus, line $d$ represents the

184: lower limit of $\dot M$ needed for NDAFs to be realized. Similar to

185: line $c$, line $e$ shows the upper limit of $\dot M$ for NDAFs,

186: beyond which the heating $Q_{\rm vis}$ is always too large to be

187: balanced by the total cooling $(Q_{\rm adv} + Q_{\nu})$, and no

188: thermal equilibrium solutions can exist. Therefore, the joining

189: point of lines $d$ and $e$ at $r \approx 185$$r_{\rm g}$, marked by

190: a filled circle, defines the maximal possible outer boundary of an

191: NDAF. We repeat that the finding of a radius-dependent maximal

192: possible accretion rate, ${\dot M}_ {\rm max}(r)$, for slim disks as

193: well as for NDAFs (i.e., lines $c$ and $e$) is because of the usage

194: of equation (1). The physical reason for the existence of ${\dot M}_

195: {\rm max}(r)$ is that the black hole's gravitational force in the

196: vertical direction, correctly calculated from equation (1), can only

197: gather some limited amount of accreted matter; beyond this limit the

198: pressure force would be too large to be balanced by the

199: gravitational force, and outflows are likely to be produced as a

200: result.

201: 

202: 

203: Note that in Figure 1 the upper limit line $c$ for slim disks and

204: the lower limit line $d$ for NDAFs join into a single line at $r

205: \thickapprox 18.6 r_{\rm g}$, and the region for slim disk solutions

206: and that for NDAF solutions combine into a single region, ranging

207: over several orders of magnitude in $\dot M$. There is no boundary

208: separating these two seemingly very different classes of black hole

209: accretion disks, slim disks and NDAFs. Both of them are optically

210: very thick for photons. As the accretion rate increases, the

211: processes of neutrino emission operate and become important, and the

212: accretion flow changes from the slim disk form to the NDAF form. By

213: also including SSD solutions, which are separated from slim disk

214: solutions by the unstable region and can possibly connect slim disk

215: solutions via limit-cycle oscillations (e.g., Szuszkiewicz \& Miller

216: 2001; Li et al. 2007), Figure 1 does provide a unified description

217: of all the three known classes of optically thick accretion disks

218: around black holes.

219: 

220: 

221: Line $e$ in Figure 1 looks to indicate that for very small radii,

222: the allowed accretion rate of NDAFs could be as high as $\sim 10^5 -

223: 10^6 M_\odot {\rm s}^{-1}$. One might worry that the density of the

224: flow matter there could reach up to the same as or even larger than

225: the nuclear density, then there would be uncertainties in

226: calculating the total pressure, the neutrino cooling rate and so on,

227: because such calculations need information about the nuclear

228: equation of state in detail which is highly model-dependent and

229: unclear at present. Our answer to this problem is that such a high

230: accretion rate for a small radius is only a theoretical upper limit

231: and is unlikely to be astrophysically realizable. All the related

232: theoretical models and numerical calculations have agreed that the

233: accretion rate needed to power a GRB is in the range of $\sim 0.01 -

234: 10 M_\odot {\rm s}^{-1}$ (e.g., Popham et al. 1999; Kawanaka \&

235: Mineshige 2007). We plot in Figure 2 the mass density on the

236: equatorial plane, $\rho_0$, as functions of $r$ for four different

237: values of $\dot{M}$: $0.01$, $0.1$, $1$, and $10 M_\odot {\rm

238: s}^{-1}$. It is seen that even for $\dot{M}= 10 M_\odot {\rm

239: s}^{-1}$, $\rho_0$ is far below the nuclear density which is $\sim

240: 10^{14}$$ {\rm g}$ ${\rm cm}^{-3}$. Our calculations also give that

241: (not drawn in Figure 2) only for an (unrealistically) high accretion

242: rate $\dot{M} \sim 10^6 M_\odot {\rm s}^{-1}$, $\rho_0$ could reach

243: up to $\sim 10^{14}$$ {\rm g}$ ${\rm cm}^{-3}$ in a very small

244: region $r \lesssim 4r_{\rm g}$. Thus, Figure 1 does not seem to

245: suffer the density problem.

246: 

247: 

248: Figure 3 shows thermal equilibrium solutions in the $\dot

249: M$-$\Sigma$ plane for fixed radii $r = 5r_{\rm g}$ (the solid line)

250: and $r = 100r_{\rm g}$ (the dashed line). Such a design is widely

251: found in the literature as it is a useful tool for the local

252: stability analysis: solutions on the lines with a positive

253: derivative $({\partial {\dot M}}/{\partial \Sigma}

254: > 0)$ are viscously stable, while those with a negative derivative

255: $({\partial {\dot M}}/{\partial \Sigma} < 0)$ are viscously unstable

256: (e.g., Kato et al. 1998). As seen from the figure, for a small

257: radius $(r = 5r_{\rm g})$ there is a continuous S-shaped sequence of

258: thermal equilibria, of which the lower and middle branches

259: correspond to stable SSDs and unstable SSDs, respectively; and the

260: upper branch extends from stable slim disk solutions to stable NDAF

261: solutions with increasing $\dot M$ and ends at the maximal possible

262: $\dot M$ for NDAFs (i.e., line $e$ in Fig. 1). For a large radius

263: $(r = 100r_{\rm g})$, however, the solution sequence is no more

264: continuous and is broken into two parts by the 'no solution' region

265: shown in Figure 1: the lower part is still for stable SSDs and

266: unstable SSDs, and the upper part is for stable NDAFs and is bounded

267: by the minimal and maximal possible $\dot M$(i.e., lines $d$ and $e$

268: in Fig. 1). Previous works (Kohri \& Mineshige 2002; Di Matteo et

269: al. 2002; Kawanaka \& Mineshige 2007) have proved that NDAFs are

270: stable. Our results here confirm this conclusion, but with an

271: additional remark that for NDAFs the allowed accretion rate has both

272: a lower and an upper limit.

273: 

274: All the above results are obtained with a one-dimensional (1-D)

275: analytical model, that is, as in most previous works on slim disks

276: as well as NDAFs (see the references in $\S 1$), we solve the

277: vertically integrated equations and do not consider the vertical

278: structure of the disk. Strictly speaking, simple 1-D models can be

279: valid only for geometrically thin disks such as SSDs. For slim disks

280: and NDAFs which are geometrically not thin, the multidimensional

281: effects on the disk's basic properties, e.g., the various timescales

282: and the cooling rates can be important or even crucial, so the

283: validity of 1-D approximations should be checked by at least

284: two-dimensional (2-D) studies. Recently, 2-D

285: radiation-hydrodynamical numerical simulations and analytical

286: treatments of supercritical accretion flows, to which slim disks and

287: NDAFs belong, have been made (Ohsuga et al. 2005; Ohsuga 2007; Kohri

288: et al. 2007). These works showed that it is indeed quite important

289: to consider carefully the vertical dependence of fluid quantities

290: and the radiative diffusion in the vertical direction. Compared with

291: the correct 2-D results, the 1-D slim disk model (Abramowicz et al.

292: 1988; Kato et al. 1998) tends to overestimate $Q_{\rm rad}$, and

293: hence, underestimate $Q_{\rm adv}$ and overestimate the luminosity.

294: Another important multidimensional effect for slim disks is that

295: photon trapping modifies the spectral energy distribution. For

296: NDAFs, however, similar multidimensional effects were only predicted

297: but not proved (Kohri et al. 2007). Having mentioned these recent

298: 2-D works, we think that the validity of our 1-D results here is

299: unlikely to be affected. We do not calculate the luminosity or the

300: radiation spectrum. What we have done is a dynamical study. We pay

301: attention to the thermal equilibrium between the viscous heating

302: $Q_{\rm vis}$ and the total cooling, which is the sum of $Q_{\rm

303: rad}$ and $Q_{\rm adv}$ for slim disks and the sum of $Q_{\nu}$ and

304: $Q_{\rm adv}$ for NDAFs. Even though $Q_{\rm rad}$ (and maybe

305: $Q_{\nu}$ too) may be overestimated and $Q_{\rm adv}$ underestimated

306: in the 1-D calculations, these two inaccuracies should tend to

307: cancel with each other, such that the resulting total cooling is

308: likely to remain a reasonable estimate. From the analysis of thermal

309: equilibrium solutions we find, as our main result, that NDAFs have

310: both a maximal and a minimal possible accretion rate at their each

311: radius. This result implies that outflows driven by the pressure

312: force are likely to originate from accretion flows with an accretion

313: rate exceeding its allowed upper limit. Such outflows have been

314: found in the 2-D simulations of supercritical accretion flows

315: (Ohsuga 2007).

316: 

317: 

318: \section{On the Origin of X-Ray Flares in Gamma-Ray Bursts}

319: 

320: 

321: We discuss briefly an astrophysical implication of our results. A

322: new, unexpected phenomenology of gamma-ray bursts (GRBs) revealed by

323: the $Swift$ satellite is erratic X-ray flares following a number of

324: long-duration and short-duration GRBs. Such flares generally rise

325: and fall rapidly, with typical time scales much shorter than the

326: epoch when the flare occurs. This time behavior strongly supports

327: the 'internal' origin of the flares, in contrast to the 'external'

328: origin of the power-law decay afterglows; that is, it indicates a

329: restart of the GRB central engine (see, e.g., Proga \& Zhang [2006]

330: for references). Although long and short GRBs are likely to be

331: associated with different types of progenitors, namely collapsars

332: for long GRBs and mergers of compact objects for short ones; it is

333: generally believed that the central engines of these two classes of

334: events are in common invoking a hyperaccretion disk around a

335: stellar-mass black hole (see, e.g., Perna et al. [2006] for

336: references). Thus, a restart of the GRB central engine is just a

337: restart of accretion. Based on these considerations, models for a

338: common origin of the X-ray flares in both long and short GRBs have

339: been recently proposed, in which accretion of a black hole is

340: repeatedly ended and then restarted either by the fragmentation in

341: the outer parts of accretion flows (Perna et al. 2006), or by the

342: magnetic flux accumulated in the vicinity of the black hole (Proga

343: \& Zhang 2006).

344: 

345: 

346: 

347: 

348:   Taking the view that hyperaccretion disks radiate their gravitational

349: energy as neutrinos to power GRBs (i.e., NDAFs, see references

350: listed in \S 1), our results here may be suggestive of a new and

351: simpler interpretation for the cause of ending and restarting the

352: accretion process. As seen from Figure 1, NDAF solutions can exist

353: only for the region bounded by lines $d$ and $e$, the joining point

354: (the filled circle) of these two limit lines defines the largest

355: possible outer boundary, $r^{\ast} \approx 185 r_{\rm g}$, of an

356: NDAF and corresponds to an accretion rate ${\dot M}^{\ast} \approx

357: 0.62 M_{\sun} {\rm s}^{-1}$. Thus, depending on the place where an

358: inflow toward the black hole originally starts and the available

359: mass supply rate, there are following three distinct cases. (1) The

360: inflow starts at a radius $r < r^{\ast}$ and the mass supply rate is

361: within the region between lines $d$ and $e$. In this case, an NDAF

362: forms and a GRB appears, but no flares follow. (2) The inflow either

363: starts at $r < r^{\ast}$ with a mass supply rate below line $d$, or

364: starts at $r > r^{\ast}$ with a mass supply rate less than ${\dot

365: M}^{\ast}$. This is the case in which no NDAFs can form and no GRBs

366: are produced at all. (3) The inflow either starts at $r < r^{\ast}$

367: with a mass supply rate above line $e$, or starts at $r > r^{\ast}$

368: with a mass supply rate larger than ${\dot M}^{\ast}$. In this case,

369: the flow must first lose matter in the form of outflows (because the

370: pressure force exceeds the gravitational force, as mentioned in \S

371: 3), such that its remaining matter ensures an accretion rate within

372: the region between lines $d$ and $e$, then there is an NDAF and a

373: GRB. This primary hyperaccretion ends when the matter in the NDAF is

374: all swallowed by the central black hole as in case 1, but this is

375: not all. Some or all of the outflowing matter can be attracted

376: backward again by the black hole. It is this feedback that is

377: responsible for restarting accretion. Such secondary accretion is

378: expected to have a rate that is substantially lower than the primary

379: hyperaccretion rate and is insufficient to support another NDAF, so

380: it is likely to be in the form of photon radiation-cooled disks that

381: produce X-ray flares. In this line, a detailed model for X-ray

382: flares in GRBs is being worked out.

383: 

384: 

385: For the moment, let us add a few more words about the realistic

386: initial conditions of hyperaccretion flows and the possibilities of

387: the mechanism for X-ray flare production suggested here, by

388: comparing the results for $\dot{M}$ and the outer boundary of NDAFs

389: given in Figure $1$ with those existed in the literature. As already

390: mentioned in $\S 3$, in all the known theoretical models and

391: numerical calculations $\dot{M}$ needed to power a GRB is in the

392: range of $\sim 0.01  -  10 M_\odot {\rm s}^{-1}$. Further, the most

393: probable $\dot{M}$ for NDAFs is in the range of $\sim 0.2  -  1

394:  M_\odot {\rm s}^{-1}$ (Kawanaka \& Mineshige 2007). For the outer

395: boundary of NDAFs, defined as the radius inside which neutrino

396: cooling is significant, Chen \& Beloborodov (2007) calculated that

397: it is $\sim 100 r_{\rm g}$ for $\alpha = 0.1$ and $\sim 200 r_{\rm

398: g}$ for $\alpha = 0.01$. All these values are consistent with the

399: NDAF region bounded by lines $d$ and $e$ in Figure $1$. Regarding

400: the realistic conditions in the progenitor systems of GRBs, it is

401: less clear where the inflow originally starts and how large the

402: initial mass supply rate is, but it seems plausible that short GRBs

403: resulted from mergers of compact objects have smaller accretion

404: disks and long GRBs resulted from collapsars have larger ones (e.g.,

405: Popham et al. 1999; Narayan et al. 2001). In view of these facts,

406: there should be possibilities for case $3$ described above to be

407: realized. For instance, if an inflow starts at $r < r^*$ with a mass

408: supply rate above line $e$, then the resulting GRB is likely a short

409: one with flares following it; or if an inflow starts at $r > r^*$

410: with a mass supply rate larger than ${\dot{M}}^*$, then it is the

411: chance for a long GRB instead, and with flares too.

412: 

413: 

414: \acknowledgments

415: 

416: 

417: We thank Z. G. Dai for beneficial discussions and the referee for

418: helpful comments. This work was supported by the National Natural

419: Science Foundation of China under grants 10503003 and 10673009 and

420: the Natural Science Foundation of Fujian Province of China under

421: grant V0750001.

422: 

423: \clearpage

424: 

425: 

426: \begin{thebibliography}{}

427: \bibitem[]{} Abramowicz, M. A., Czerny, B., Lasota, J.-P., \& Szuszkiewicz, E. 1988, \apj, 332, 646

428: \bibitem[]{} Chen, W.-X., \& Beloborodov, A. M. 2007, \apj, 657, 383

429: \bibitem[]{} Di Matteo, T., Perna, R., \& Narayan, R. 2002, \apj, 579, 706

430: \bibitem[]{} Gu, W.-M., Liu, T., \& Lu, J.-F. 2006, \apj, 643, L87

431: \bibitem[]{} Gu, W.-M., \& Lu, J.-F. 2007, \apj, 660, 541

432: \bibitem[]{} H\={o}shi, R. 1977, Prog. Theor. Phys., 58, 1191

433: \bibitem[]{} Janiuk, A., Yuan, Y.-F., Perna, R., \& Di Matteo, T. 2007, \apj, 664, 1011

434: \bibitem[]{} Kato, S., Fukue, J., \& Mineshige, S. 1998, Black-Hole

435: Accretion Disks (Kyoto: Kyoto Univ. Press)

436: \bibitem[]{} Kawanaka, N., \& Mineshige,S. 2007, \apj, 662, 1156

437: \bibitem[]{} Kohri, K., \& Mineshige, S. 2002, \apj, 577, 311

438: \bibitem[]{} Kohri, K., Narayan, R. \& Piran, T. 2005, \apj, 629, 341

439: \bibitem[]{} Kohri, K., Ohsuga, K., \& Narayan, R. 2007, \mnras, 381, 1267

440: \bibitem[]{} Lee, W. H., Ramirez-Ruiz, E., \& Page, D. 2005, \apj, 632, 421

441: \bibitem[]{} Li, S.-L., Xue, L., \& Lu, J.-F. 2007, \apj, 666, 368

442: \bibitem[]{} Liu, T., Gu, W.-M., Xue, L., \& Lu J.-F. 2007, \apj, 661, 1025

443: \bibitem[]{} Narayan, R., Piran, T., \& Kumar, P. 2001, \apj, 557, 949

444: \bibitem[]{} Ohsuga, K. 2007, \apj, 659, 205

445: \bibitem[]{} Ohsuga, K., Mori, M., Nakamoto, T., \&Mineshige, M. 2005, \apj, 628, 368

446: \bibitem[]{} Paczy\'nski, B., \& Wiita, P. J. 1980, \aap, 88, 23

447: \bibitem[]{} Perna, R., Armitage, P. J., \& Zhang, B. 2006, \apj, 636, L29

448: \bibitem[]{} Popham, R., Woosley, S. E., \& Fryer, C. 1999, \apj, 518, 356

449: \bibitem[]{} Proga, D., \& Zhang, B. 2006, \mnras, 370, L61

450: \bibitem[]{} Shakura, N. I., \& Sunyaev, R. A. 1973, \aap, 24, 337

451: \bibitem[]{} Szuszkiewicz, E., \& Miller, J. C. 2001, \mnras, 328, 36

452: 

453: 

454: 

455: \end{thebibliography}

456: 

457: \clearpage

458: 

459: \begin{figure}

460: \plotone{f1.eps} \caption{ Distribution of black hole optically

461: thick accretion disk solutions. The $\dot{M}$-$r$ plane is divided

462: into four regions by five boundary lines $a$, $b$, $c$, $d$, and

463: $e$. The filled circle denotes the largest possible outer boundary

464: of an NDAF. \label{fig1}}

465: \end{figure}

466: \clearpage

467: 

468: 

469: \begin{figure}

470: \plotone{f2.eps} \caption{Radial dependence of the mass density

471: $\rho_0$ of NDAFs for different possible values of $\dot{M}$: $0.01$

472: (dashed line), $0.1$ (dotted line), $1$ (solid line), and $10$

473: (dot-dashed line) $M_\odot {\rm s}^{-1}$. \label{fig2}}

474: \end{figure}

475: \clearpage

476: 

477: 

478: \begin{figure}

479: \plotone{f3.eps} \caption{Stability curves in the $\dot{M}$-$\Sigma$

480: plane for radii $5 r_{\rm g}$ (solid line) and $100 r_{\rm g}$

481: (dashed line). \label{fig3}}

482: \end{figure}

483: 

484: \clearpage

485: 

486: 

487: \end{document}

488: